The anti-competitive effects of pricing algorithms: some thoughts on algorithmic cartels
p. 487-497
Texte intégral
1The law of Allarde of March 2 and 17, 17911 provides for the possibility for any person “to do such trade or to exercise such profession, art or trade as he finds good”, a principle reinforced by the law of June 14 and 17, 1791, known as the Le Chapelier law2, which abolished corporations. More simply, these two texts founded the principle of freedom of trade and industry, which is the fundamental principle according to which any natural or legal person can carry out a commercial activity of his choice and exploit it freely3. A public entity cannot intervene, in any way, to distort the game of competition which must be free and fair4.
2However, this freedom of competition remains limited: competition must not disturb the free course of a market. It is in this context of guaranteeing free competition without undermining the activity carried out in this market that the reflections on the use of new technologies linked to artificial intelligence within the markets were conducted. Our legal rules and our economy have undergone a real upheaval with the emergence and development of new digital technologies, especially with artificial intelligence, which constitutes a formidable tool for the development of competition5.
3Among the digital tools powered by artificial intelligence, pricing algorithms are controversial advances, making companies more innovative and productive. But their use for collusive purposes is forbidden. For example, in 2015, the founder of the company Posters Revolution was prosecuted for participating in a cartel between different competitors to influence the prices of online product sales6. In fact, they can facilitate agreements between different competitors. Indeed, they allow price alignment and market surveillance practices between them7. Once companies have agreed to create these algorithms, the elements constituting an anti-competitive practice are brought together8. Price algorithms are becoming more and more important for companies in their quest for innovation and efficiency, and it is therefore necessary to know their effects on the functioning of competition within the markets9. This study aims to reveal the different categories of pricing algorithms in order to understand their functioning and purpose (I), and the consequent infringements of competition law (II).
I. Overview of the different categories of pricing algorithms
4Although there is no legal definition of a price algorithm, or even of an algorithm at all (A), it is possible today to establish a classification of different price algorithms (B).
A. The absence of a legal definition of the pricing algorithm
5There is no legal definition of an algorithm, let alone a price algorithm. The notion is old, “and much older than the creation of the first computers; there is no universally accepted definition”10. In the current sense of the term, the algorithm is “the description of a sequence of steps making it possible to obtain a result from elements provided in input”11, expressed in a computer language. In other words, it is a sequence of instructions intended to solve a problem through the implementation of various computer operations. The objective of the algorithm is to be as efficient as possible: what defines the level of correctness of the algorithm is its ability to solve a defined problem12. It is written in what is called a programming language13.
6The French Competition Authority attempted to define the pricing algorithm in a March 6, 2018 recommendation on data mining in the internet advertising sector14. The algorithm “means a series of rules to be applied in a specific order to accomplish a particular task: it is a logical sequence to obtain a certain result from a given input”15. The price algorithm will have consequences in an economic market, in particular by contributing to the setting of prices16.
B. Presentation of the different pricing algorithms
7Today, it is possible to classify pricing algorithms into two categories: descriptive algorithms, and black box algorithms that work through artificial intelligence.
8The descriptive algorithm is programmed to follow a predefined strategy. Its objective is to monitor the prices used by competing merchants in the marketplaces. Thanks to this increased monitoring of the competition, it can automatically adjust the prices offered17. These descriptive algorithms observe and evaluate the market by analysing the data in a first step, and choose the most optimal price in a second step. They set the price based on various other factors, such as data obtained on the purchasing power of customers, their wishes, especially through cookies or past purchases18. With these features, they can then adjust the prices.
9The black box algorithm is much more sophisticated since it works thanks to machine learning. Machine learning is an artificial intelligence technology that allows the algorithm to learn from its environment and to act almost autonomously. These algorithms that work thanks to machine learning are called self-learning algorithms and differ from the more classical algorithms which work thanks to more summary information provided by the operators19.
10The use of pricing algorithms, whatever their nature, has effects on competition. It is important to monitor their use because they can be the source of anticompetitive practices, but the benefits they provide are no less important.
II. The anti-competitive effects of the use of algorithms: the example of the algorithmic cartel
11Price algorithms produce anti-competitive effects (A), but we cannot ignore the interesting and beneficial effects they produce, particularly in terms of profitability and attractiveness for companies and consumers (B).
A. Pricing algorithms and cartels: examples of anti-competitive practices
12Competition allows companies to be in harmony in the same market in order to guarantee a form of economic progress. This is why anti-competitive practices will produce harmful consequences in this market, which justifies their prohibition. Cartels are among the prohibited anti-competitive practices. Under domestic law, Article L. 420-1 of the French Commercial Code prohibits all practices that could have the effect of preventing, restricting or distorting competition in a given market. In European law, article 101 of the Treaty on the Functioning of the European Union (TFEU) provides that “the following shall be prohibited as incompatible with the internal market: all agreements between undertakings, decisions by associations of undertakings and concerted practices which may affect trade between Member States, and which have as their object or effect the prevention, restriction or distortion of competition within the internal market” 20. A cartel exists, for example, when companies agree to make higher profits than they should in a situation of normal competition. To be constituted, the cartel is conditioned to the existence of a meeting of the consents of the parties to the cartel21, and a concertation between these parties which restricts competition on the market22.
13First of all, concerning the existence of a concerted action: to exist, this concerted action supposes a joint consent of the parties whose will is to restrict competition23, it is therefore necessary to demonstrate the intention of the actor to join this action. It supposes a collective behaviour24. Article 101 of the Treaty on the Functioning of the European Union and L. 420-1 of the Commercial Code refers to “concerted actions, agreements, express or tacit understandings”25, which means that concerted actions can take different forms. As far as agreements are concerned, their form is not important26. If it is a concerted practice, it does not necessarily imply the demonstration of the existence of a strategy27. This agreement is concluded between companies operating in a common market, supposed to be in competition with each other, regardless of their place in this market28. Article L. 410-1 of the Commercial Code refers to both express and tacit agreements.
14Secondly, this concertation must produce anti-competitive effects. This means that competition on the market is disturbed in at least one parameter contributing to its balance. The agreement constituting the cartel must have as a consequence or as an object to harm competition29. Thus, the competitive effects will not be of the same level from one market to another. The assessment of these anti-competitive effects is the work of case law. The Court of Justice of the European Union has identified ways to achieve an infringement. They are based either on the “alterations in competition”30 which “must result from all or part of the clauses of the agreement itself”31, or on the analysis of the effects of the agreement on competition32, or on the analysis of the effects of the agreement. In other words, competition may be directly affected by the existence of the agreement, or it may be affected by the implementation of the agreement. This means that a cartel may be anti-competitive even though the agreement has not been implemented33.
15The pricing algorithm can be used to promote a cartel. Competitors will agree to use an algorithm in a common way and will be able to adopt the same prices. The algorithm can also be used to ensure a form of stability of the cartel by monitoring the operators who take part in the cartel. It is in similar circumstances that the European Commission sanctioned operators who controlled the prices charged by their distributors. Algorithms were used to monitor prices. Thanks to this monitoring, the leading operator could automatically adapt the prices charged by its distributors to those of competing distributors34.
16Pricing algorithms can also contribute to the formation of price cartels35. The French Commercial Code provides for the principle of the prohibition of agreements regardless of the form they may take36. The Treaty on the Functioning of the European Union prohibits agreements that could affect the market between Member States: “The following shall be prohibited as incompatible with the internal market: all agreements between undertakings, decisions by associations of undertakings and concerted practices which may affect trade between Member States, and which have as their object or effect the prevention, restriction or distortion of competition within the internal market”37.
17A report issued by the French and German competition authorities details the various cartel scenarios that could arise from the use of pricing algorithms38. In very general terms, cartelisation is defined as “a situation in which firms use reward and punishment schemes to reward competitors when they engage in behaviour leading to supra-competitive outcomes and punish them when they deviate”39. In such a context, the use of pricing algorithms can be a means of monitoring and sanctioning such firms within a distribution network. Indeed, a supplier is equipped with the necessary technology to monitor and control the prices charged by its resellers through tracking algorithms40.
18Another situation envisaged by the competition authorities’ report is the case where several firms will use the same pricing algorithm, which is provided by a third party. These firms do not necessarily work together, yet this pricing algorithm will be used in the pricing decision. This will lead to an inevitable alignment of prices, even if the companies did not agree to this agreement in advance41. Here is an example of applications that allow individuals to book private drivers or chauffeur-driven cars: the platform, thanks to an algorithm, will put the driver in contact with the customers. The prices are then standardised without the drivers having been able to agree among themselves on these prices42.
19Pricing algorithms can also allow competing companies to tacitly align their business strategies. Thanks to machine learning, the algorithms used by each company will be able to learn from their environment without any human intervention. The algorithms will collect data about the competitors’ business practices, which will allow the companies to align their prices with them. Such an operation reduces to nothing any incentive to compete. Let us remember how machine learning works: autonomous systems capable of learning and adapting without following direct instructions43. The progress of artificial intelligence is such that it would be quite possible for price coordination to take place without any human decision-making44.
20Finally, pricing algorithms can be used to hide the possible presence of a cartel in a market. The Topkins case revealed the use of these algorithms45. In this case, a poster seller on Amazon agreed with its competitors on the use of a price algorithm that would allow it to be adjusted in the event of the occurrence of an external event that would lead to price differences between them46. This case demonstrated the implementation of an agreement between the companies involved; however, the particularity of this agreement was to rely on a price algorithm.
21Competition infringements are not only about cartels. For example, the European Commission fined Google €2.42 billion for non-compliance with competition rules47. Google took advantage of its dominant position as a search engine and gave an illegal advantage to its price comparison service. The European Commission ordered Google to put an end to this practice within 90 days or face a penalty payment of up to 5% of average daily turnover48. The abuse of a dominant position has been defined by jurisprudence as any “power to hinder the maintenance of effective competition on the relevant market by providing the possibility of independent behaviour to an appreciable extent vis-à-vis its competitors, its customers and ultimately consumers”49. Article 102 of the Treaty on the Functioning of the European Union prohibits “the abuse by one or more undertakings of a dominant position within the internal market or in a substantial part of it”50.
B. The beneficial effects of the use of pricing algorithms: a tool to strengthen the competitiveness of companies?
22A pricing algorithm will adjust the price according to supply and demand in order to offer consumers the fairest price possible51.
23Pricing algorithms also play a role in companies’ business strategies. Indeed, their use combined with other technologies such as artificial intelligence has beneficial consequences for economic markets. Indeed, they can perform important and complex tasks, such as data collection52. Moreover, thanks to the development of machine learning, it has become possible for machines to learn by themselves: indeed, some algorithms have a high degree of autonomy that allows them to adapt and evolve their parameters without any human intervention53.
24Companies are obviously interested in pricing algorithms. Indeed, their use helps workers because they perform different tasks, and they allow to better understand the market and thus to adapt the practices carried out by companies. However, these digital innovations do not present great risks. On the other hand, pricing algorithms present more worrying aspects in that they can undermine markets and competition law when their purpose is to carry out a strategy aimed at crowding out competitors on the market54.
25Pricing algorithms are among the new technologies of the digital sector that contribute to make companies more competitive and innovative, and therefore, more efficient. But questions remain concerning their use for purposes constituting anti-competitive behaviour, and concerning the constitution of algorithmic agreements. The competition authorities are competent to investigate. However, it seems necessary to provide them with the tools they need to better detect these competitive practices. Furthermore, a stricter framework for the use of pricing algorithms, including a legal definition and rules of good practice regarding their use, would make it possible to reverse the trend by making them more profitable for companies. The price algorithm would become a real tool for the promotion of competition in the markets.
Notes de bas de page
1 Loi du 2-17 mars 1791, portant suppression de tous les droits d’aides, de toutes les maîtrises et jurandes et établissement des droits de patente.
2 Décret du 14 juin 1791 relatif aux assemblées d’ouvriers et artisans de même état et profession, dit « Loi Le Chapelier », Recueil Duvergier, p. 25-26.
3 O. Moussa, « Parasitisme et concurrence dans les mondes virtuels », in G. Delabre (dir.), Le droit dans les mondes virtuels. Philosophie et économie : Larcier, Coll. du CRIDS, 2019, p. 138.
4 G. Clamour et P.-Y. Gahdoun, « Commerce et industrie », Rep. dr. commercial, fév. 2019, §47 et suiv. (updated : july 2021).
5 T. Titone, « Intelligence artificielle et droit de la concurrence », RLDA, 2019, n° 151, p. 1.
6 United States of America, District Court, Northern District of California, San Francisco Division, United States of America v. David Topkins, April 30, 2015.
7 AUTORITÉ DE LA CONCURRENCE (fr) and BUNDESKARTELAMT (de), Algorithms and competition [on line], Paris/Bonn, Nov. 2019, 92 pages. Available at: https://www.autoritedelaconcurrence.fr/sites/default/files/Algorithms-and-competition_FR.pdf (accessed on August 18, 2021).
8 Ibid.
9 L. Arcelin, « Le droit de la concurrence mis à l’épreuve par le numérique », JCP E, 2019, n° 47, p. 1493.
10 OCDE, DIRECTION DES AFFAIRES FINANCIÈRES ET DES ENTREPRISES COMITÉ DE LA CONCURRENCE, Algorithmes et ententes. Note d’information au Secrétariat [on line], DAF/COMP(3017)4, June 16, 2017, p. 6. Available at: https://one.oecd.org/document/DAF/COMP(2017)4/fr/pdf (accessed on August 18, 2021). Personal translation.
11 CNIL, Algorithme [on line]. Available on: https://www.cnil.fr/fr/definition/algorithme (accessed on August 18, 2021).
12 O. Houdé, « L’intelligence humaine n’est pas un algorithme », Éd. Odile Jacob, p. 45.
13 Ibid.
14 AUTORITÉ DE LA CONCURRENCE, « avis n° 18-A-03 portant sur l’exploitation des données dans le secteur de la publicité sur internet » [on line], March 6, 2018, 125 pages. Available at: https://www.autoritedelaconcurrence.fr/fr/avis/portant-sur-lexploitation-des-donnees-dans-le-secteur-de-la-publicite-sur-internet (accessed on August 22, 2021).
15 AUTORITÉ DE LA CONCURRENCE, avis n° 18-1-6, op. cit., p. 116. An input can be defined as the raw material that contributes to the manufacture of something. It is an integral part of the finished product.
16 F. Marty, « Algorithmes de prix, intelligence artificielle et équilibres collusifs », Revue internationale de droit économique, 2017, p. 4. Also see : F. Rouvière, « Jus ex machina : la normativité de l’intelligence artificielle », RTD civ. 2019, p. 217.
17 F. Marty, op. cit.
18 M. Malaurie-Vignal, « Concurrence et numérique : un foisonnement d’idées pour dominer les géants », Comm., com., électr, 2020, étude n° 17, p. 3.
19 Ibid.
20 Also see : J.-B. Blaise, « Entente », Rep. dr. européen, janv. 2016, §22, (updated : sept. 2017).
21 « Entente » in J.-P. SCARANO (dir.), Dictionnaire de droit du marché. Concurrence, distribution, consommation : Ellipses, coll. Dictionnaires de Droit, 2008, p. 161.
22 Y. Auguet, C. Galoko et A. Riéra, Droit de la concurrence : Ellipses, coll. Spécial Droit, 2020, p. 145.
23 C. Bolze, S. Retterer, C. Hendericksen, F. Thibault, « Règles de concurrence. Information de la commission européenne relative à sa politique en matière d’amendes pour infraction aux règles de concurrence », RTD com. 1996, p. 587 ; E. Claudel, « Éléments constitutifs de l’entente », RTD Com. 2003, p. 495 ; M. Malaurie-Vignal, « Droit de la concurrence et droit des contrats », D. 1995, p. 51.
24 Ibid. Also see : M. Picard, « Accord de volontés », Rep. cont. Adm., apr. 2005, §77 (updated : apr. 2017) ; E. Claudel, « Rapport annuel 2006 du Conseil de la concurrence », RTD com. 2007, p. 698.
25 Personal translation. C. com., art. L. 420-1 : « sont prohibées même par l’intermédiaire direct ou indirect d’une société du groupe implantée hors de France, lorsqu’elles ont pour objet ou peuvent avoir pour effet d’empêcher, de restreindre ou de fausser le jeu de la concurrence sur un marché, les actions concertées, conventions, ententes expresses ou tacites ou coalitions »
26 Comm. CE, déc. 94/599/CE, 27 juillet 1994, PVC, pt. 30 ; Comm. CE, déc. 2004/138/CE, 11 juin 2002, Club Lombard, pt. 412. Also see : TPICE, 15 mars 2000, aff. T-25/95, Cimenteries CBR/Commission.
27 Y. Auguet, C. Galoko et A. Riéra, op cit., p. 146.
28 CJCE, 30 juin 1996, LTM c/ MBU, op. cit.
29 A. Decocq et G. Decocq, Droit de la concurrence. Droit interne et droit de l’Union européenne : LGDJ, 8ème éd., coll. Manuel, 2018, p. 304.
30 CJCE, 30 juin 1996, LTM c/ MBU, aff. 56/65, R, 337, concl. Av. gén. K. Roemer. Personal translation.
31 Ibid. Personal translation.
32 About the proof of the fraudulent agreement : J.-C. RODA, « Le standard de preuve : réflexions à partir du droit de la concurrence », D. 2021, p. 1297 ; S. Atsarias-Dumas, « Le contrôle du contrat par l’Autorité de la concurrence », AJ contrat 2020, p. 474.
33 TUE, 8 sept. 2016, aff. T-472-13 Lunbeck c/ Commission, n° 341.
34 As an example : Comm. UE, 24 juill. 2018, aff. AT. 40465, Asus : C (2018)4773.
35 L. Arcelin, « Le droit de la concurrence mis à l’épreuve par le numérique », JCPE E, n° 45, 2020, p. 1493.
36 C. com., art. L. 420-1.
37 TFUE, art. 101.
38 AUTORITÉ DE LA CONCURRENCE (fr) and BUNDESKARTELAMT (de), Algorithms and competition [on line], Paris/Bonn, Nov. 2019, 92 pages. Available at: https://www.autoritedelaconcurrence.fr/sites/default/files/Algorithms-and-competition_FR.pdf (accessed on August 18, 2021). Also see : S. Troussard et F. De Bure, « La coordination algorithmique : fantasme ou réalité ? », RLDA, 2020, n° 92.
39 AUTORITÉ DE LA CONCURRENCE, op. cit., p. 4. Personal translation.
40 J.-C. Roda, « L’entente algorithmique », JCP G, 2019, n° 28, doctr. 785, p. 1373.
41 OECD, Directorate for financial and enterprise affaires competition committee, « Algorithms and Collusion » note by the United States », DAF/COMP(2017)4, juin 2017, p. 7.
42 S. Troussard et F. De Bure, op. cit.
43 About the machine learning : L. Viaut, « Droit et algorithmes : réflexion sur les nouveaux processus décisionnels », LPA, 2020, n° 177-178, p. 8 ; C. Crichton, « Projet de règlement sur l’IA (I) : des concepts larges retenus par la Commission », Dalloz actualité, 3 mai 2021.
44 J.-C. Roda, op. cit.
45 United States of America, District Court, Northern District of California, San Francisco Division, United States of America v. David Topkins, April 30, 2015.
46 F. Marty, « Algorithmes de prix, intelligence artificielle et équilibres collusifs », Revue internationale de droit économique, 2017, p. 84.
47 Comm. UE, 27 juin 2017 aff. AT.39740, Google Search Shopping. Also see : C. Prieto, « Nouveaux abus de position dominante de Google : après celui lié à Google Shopping, ceux relatifs à Android », RTD eur. 2018, p. 513 ; R. Amaro, « Abus de position dominante et compétence internationale ou l’art subtil du précédent », D. 2021, p. 1064 ; A.-S. Choné-Grimaldi, « Google coupable d’abus de position dominante : les carottes sont-elles cuites ? », D. 2015, p. 2451.
48 Ibid.
49 CJCE, 13 fév. 1979, Hoffmann-La Roche & Co. AG, aff. C- 85/76, pt. 41.
50 TFUE, art. 102.
51 T. Titone, « Intelligence artificielle et droit de la concurrence », RLDA, 2019, n° 151, p. 2.
52 O. Sautel, « Personnalisation tarifaire à l’heure des big data : quel éclairage de la théorie économique ? », Concurrences, n° 4, novembre 2017, p. 21-24.
53 S. Troussard et F. De Bure, « La coordination algorithmique : fantasme ou réalité ? », RLDA, mars 2020, n° 92.
54 M. Malaurie-Vignal, « Concurrence et numérique : un foisonnement d’idées pour dominer les géants », Comm., com., électr, 2020, étude n° 17, p. 3.
Auteur
Doctorante en droit privé
Université Toulouse 1 Capitole
Centre de droit des affaires
Le texte seul est utilisable sous licence Licence OpenEdition Books. Les autres éléments (illustrations, fichiers annexes importés) sont « Tous droits réservés », sauf mention contraire.
Qu’en est-il du code du commerce 200 ans après ?
États des lieux et projections
Corinne Saint-Alary-Houin (dir.)
2008
Qu'en est-il de la simplification du droit ?
Frédérique Rueda et Jacqueline Pousson-Petit (dir.)
2010
La réorientation européenne de la TVA à la suite du renoncement au régime définitif
Francis Querol (dir.)
2014
Regards critiques sur quelques (r)évolutions récentes du droit
Tome 1 : Bilans et Tome 2 : Réformes-Révolutions
Maryvonne Hecquard-Théron et Jacques Krynen (dir.)
2005